# Opus 5.5 launched on day 5; a list of 'tools for having Claude make videos' hit 1,600 stars: everyone is collecting works, it collects tools · GitHub open-source projects

> Source: https://stringers.io/en/github/gh_zhuyansen__awesome-claude-video-skills.html (Stringers / 什么值得做, generated 2026-11-08)
> Stringers answers one question every day: what small product can one or two people build in a few months and actually make money from. Numbers come from 15 databases we collect daily, each with source and date.
> For agents: these verdicts and the underlying databases (verified revenue, App Store / Google Play charts, ad libraries, GitHub, Product Hunt…) are queryable live over MCP — tools today / hot / products / directions / direction / data. Setup: https://github.com/komako-workshop/stringers (the user logs in at stringers.io → "配置 agent" to get a personal URL). Site index: https://stringers.io/llms.txt

> After Opus 5.5 launched, "making video with AI" flooded the feed. Others were collecting works and prompts; it was the first to collect "which tools to install." The author posted long-form X threads in the Chinese AI circle, several big accounts reposted within an hour, and later steady star growth came from ranking first in GitHub and Google search.

[GitHub open-source project](https://stringers.io/en/github/index.md) · written 11-6

zhuyansen/awesome-claude-video-skills · zhuyansen

# Opus 5.5 launched on day 5; a list of 'tools for having Claude make videos' hit 1,600 stars: everyone is collecting works, it collects tools

**What it is** A GitHub list collecting open-source skills and tools that let coding agents like Claude Code and Codex make and edit videos, categorized, with safety levels marked, plus a real test of 16 skills making the same clip.

**Why it took off** After Opus 5.5 launched, 'AI video making' flooded feeds; others were collecting works and prompts, while it was the first to collect 'which tools to install'. The author posted a long Chinese post on X in the Chinese AI circle, several big accounts reposted it within an hour, and later GitHub and Google search kept it ranking first, with stars growing steadily.

**B** **How you could follow** Stop making same-topic lists; around the next major model or platform launch, within a week make a directory of 'what tools to use for X', with grades and a real test including a 'no tools installed' control. Run both a GitHub repo and an SEO web page, auto-update the README daily, and coordinate with friendly big accounts for the launch push.

Grade B: the playbook is clear and can be copied to niches (the AI short-drama version using the same approach got 1,901 stars), but this topic is already taken, it bets on timing the next hot event, and you need a circle that can push it within an hour; the author's own 20 other same-template lists only took off once. Word in small circles

Data (updated daily) · Repo created 2026-09-27 · 1,705 stars · 7 days +348

GitHub · stars

Data 2026-10-09 → 2026-11-08 (31 days, 31 observations). Points before 10-28 come from deeply / archive backfill, not every day has one

awesome-claude-video-skills is a GitHub list of open-source skills and tools that let coding agents like Claude Code and Codex make, edit, and animate video. It has 427 entries, each with a safety rating, in both Chinese and English. It was created in Sep 2026 and got 27 stars in 40 days (GitHub daily star count, 1,617 11-06), and it's still adding fifty to sixty a day. Three things drove the growth: the 「make videos with AI」 wave around Opus 5.5, the author's network on X in the Chinese AI community, and search traffic that came later. If you want stars for your own open-source project, the thing to copy is the method, not this topic.

### a long post titled "107 hidden gems under 50 stars"

2026 Sep 22 Opus 5.5 launched, and 「Opus 5.5 animated videos」 flooded X for two or three days. On day 5, Sep 27, a little after 10 pm Beijing time, Jason Zhu (@GoSailGlobal), who does promotion for brands going global, posted a long Chinese thread on X titled 《107 hidden gems with under 50 stars: I categorized and open-sourced all 180 open-source skills that let Claude make videos》.

The opening line sums it up: 「180 repos, 10 categories, and every one has been read through its README and given a safety rating.」 He then explains the cost: the whole process came to about $200, with a model reading each README to classify and rate it. The distribution table follows: editing and post-production 35, explainers and education 32, frameworks 29, product promotion 26, and so on. Digital humans have only 4, and music videos only 2. There's also a number that makes you want to click: of the 180, 107 have under 50 stars, and only 20 are over a thousand. 「Star-sorted leaderboards will miss all of them.」

Over an hour later, Tiejiang (@lxfater, about 109k followers) picked 4 from it and posted: 'Now Opus 5.5 plus the right skills can stand in for a junior editor': cut-video for rough cuts, and AdobePremiereProMCP, which can take over Premiere, for fine edits, one for subtitles and one for output. The last line was 'For the rest, reading his article is enough'. Cang He (@canghe) reposted within half an hour too: 'This one is 🐂🍺, bookmarked to study'. The replies were mostly 'bookmarked directly' and 'going to run the repo tonight'.

The next day the repo gained 147 stars, and on the third day 108 stars (Trendshift daily data).

### What it actually is

One README, plus a web page with filters (agentskillshub.top/best/claude-video-skills/). Each README row is one repo: star count, one-line description, safety level. Divided into 10 categories: frameworks, product promotion, explainers, editing, short video, digital humans, story animation, motion graphics, scripts, and music videos.

Compared with a typical awesome list, it adds three things:

- **Inclusion criteria are stated.** It must cover making videos, editing them, or creating motion graphics; it must be built for agents (skill, plugin, MCP, or toolkit); it must have a README; repos under 50 stars must also pass a quality bar (a finished showcase, and a one-command setup).

- **Safety rating.** The first version has 178 SAFE and 2 CAUTION, rated on the README and install steps.

- **A same-prompt test.** On Oct 11, the author had 16 of these skills each build the same 15-second promo video, and also ran plain Claude Code with nothing installed. The verdict is candid: all 16 videos passed, and the bare setup passed too, taking just 4.4 minutes. The differences were speed (fastest 2 min, slowest 23.5 min), audio (9 of 16 had sound), and look (6 looked about the same as the no-skill video). In the end he only pushed 3: explainroo, onetake, and Remotion's official skills.

Maintenance takes almost no effort. The site behind it refreshes every 8 hours, the README is regenerated automatically every day, and the list grew from 180 entries in the first version to 427. The whole repo has only 2 contributors, and the merged PRs are 0.

### Where the stars came from

Star growth came in two phases (daily star counts from Trendshift and GitHub):

| Timeline | Daily new installs | Source |
|---|---|---|
| 9-27 to 9-29 | +30, +147, +108 | Long-form posts and big Chinese accounts reposting |
| 9-30 to 10-24 | +12 to +35 | Plateau |
| After 10-25 | +35 to +84, with a single-day peak of 81 (10-27) | Search and an English content wave |
| last 7 days / previous 7 days | +404 / +398 | A flat curve with no spikes |

**Phase one ran on the author's own channels.** The author had posted 14k+ tweets on X, and this repo was built for brand promotion. Within an hour of the long post, several Chinese AI accounts followed up together, which looks like a circle of mutual boosters. He kept posting afterward: on Sep 28 he wrote "204 video-making skills packaged for use," and on Oct 21 he posted the test as "16 video skills, each building the same 15-second promo. All passed. Plain Claude Code passed too," with links to both the site and GitHub.

**Phase two ran on search.** In late October, a wave of English LinkedIn posts and blogs titled "N GitHub repos that turn Claude into a video studio" appeared, and one blog about Claude video skills directly cited this repo's test. Searching "claude video skills" on GitHub, it ranks first by stars, with the second place at only 17 stars (11-06 tested). The companion site's title, "Claude Video Editing & Remotion Skills: 431 Graded (2026)," also ranks high on Google. The repo's topics are stuffed with trending terms like claude-code, remotion, hyperframes, and opus-5-5. There were no new big-account reposts, yet it kept gaining five or six dozen stars a day, a shape that looks like search traffic (an inference; no exact breakdown is available).

**There's also a small feedback loop.** The repo has 12 issues, mostly skill authors requesting "Add: my repo." People building video skills started treating "get into this list" as a distribution channel, and the more complete the list, the more people submit.

One more thing to be clear about: the repo was less than 60 days old on GitHub and never reached trending, so these stars did not come from trending. GitHub doesn't expose the account-level breakdown of who starred it, so star-farming can't be ruled out entirely, but the shape of event peak, plateau, then steady climb looks like organic growth.

### Who else is doing the same thing

| Project | Scale (early November) | Compared with alternatives |
|---|---|---|
| yihui-dev/awesome-opus5-5-videos | 13,865 stars | Created the same day, but it collects videos and prompts made with Opus 5.5 for people to watch; it collects tools to install for agents |
| opusvideo/awesome-claude-video and six or seven other showcase / prompt collections | 36 to 238 stars | Created the same week, all showcases and prompts, with no tool directory, grading, or testing |
| The author's own awesome-opus-5.5-video | 307 stars | Same day, same person, a showcase collection with one-fifth the stars |
| ComposioHQ, travisvn, and other awesome-claude-skills | 10k to 79k stars | Collects every kind of skill, with video taking up only a few lines |
| remotion-dev/skills | 6,409 stars | A single skill, not a directory; this repo included it and ranked it third in its test |
| heygen-com/hyperframes-community-skills | 251 stars | Covers only HyperFrames |
| frankxai/awesome-motion-design-agent-skills | 97 stars | Existed since June, leans toward web motion effects, with no grading, testing, or promotion |

In the same wave, the largest one (13.9k stars) collects showcases. The next question for someone who has watched them is "what should I install?" and this is the only layer doing that.

The most telling part is the author's own control group. Using the same template, he batch-created 20+ awesome lists: codex, design, SEO, Obsidian, images, Telegram bots, hooks... Besides the video one, the best performers reached 60 and 41 stars, while most of the rest landed at 0 to 11 stars. Same template, same author, similar promotion, and only the one that caught the Opus 5.5 video craze took off.

### People who learned from it

- **awesome-short-drama-skills** (created Oct 13, 1,901 stars): the same approach of "curated niche skills + Chinese + auto-generated candidates," moved to AI short dramas and comics, and already has more stars than this one. Not verified whether it was copied from this repo.

- **dsh-video-skills** (29,2 stars in October): a plugin for "browsing 373 video skills and installing them," with a data scale close to this one's.

- A large batch of same-named xxx/claude-video-skills repos, mostly at 0 to 17 stars, mostly skill bundles rather than directories.

The short-drama list shows the approach still works in a different niche; the cluster of same-named repos shows that making another one on the same topic gets no attention.

### If it were you

**In one line:** this topic is already taken, so don't build another "Claude video skills list"; what's worth learning is the playbook: wait for the next big event and build the directory of "what tool to use next."

**Who it's for:** anyone with a circle that can get moving within an hour (a few Chinese or English AI accounts), who can script automatic collection, grading, and daily README updates.

**Who it's not for:** people with no channel to speak through, who expect "if the list is good, people will come." Of the author's 20+ lists, only one took off; the ones that missed the moment sit at single-digit stars.

**How to do it:**

1. Watch the big launches from models and platforms (new models, new agent capabilities, new platform openings). Two or three days after a launch, people collect the showcases and prompts first; you collect "tools to do the thing": skills, MCP servers, plugins.

2. State the inclusion criteria clearly, read each README and assign a safety level, and ship a first version with a hundred or two entries, sorted into categories.

3. Run a same-prompt test, include "no tools installed" as a control, and report the results straight. A test is easier to share than a list; this repo's second post and the English blog citations were both about the test.

4. Launch a GitHub repo and an independent site together. Title the site "XX Skills: N Graded (2026)" to claim the search terms; auto-update the README daily; open a "Add: my repo" submission channel so the listed authors repost for you.

5. Switching audiences also works: this one started in Chinese and is maintained bilingually, and similar directories in English, Japanese, and Korean communities are worth a look (inferred; not checked one by one).

**How to get users:** on day one, long posts plus concentrated reposts from big accounts within an hour; the first-day peak sets your starting point in GitHub's internal search. After that, rely on keyword coverage and being cited during English content waves, then submit to large awesome lists with tens of thousands of stars, like awesome-claude-code.

**When to be early:** within a week of the event. This repo launched on day 5 after the release, and six or seven similar ones appeared the same week; wait two weeks and the top search spot belongs to someone else.

**Signs you've misread it:** after the first-day reposts, daily additions drop to single digits within a week, and your repo doesn't make the top three when searching your keywords on GitHub. That means either the topic doesn't match demand or the spot is already taken. Move to the next event early instead of holding on.

[GitHub repo ↗](https://github.com/zhuyansen/awesome-claude-video-skills)

**Research notes: competitor table, acquisition evidence, full research report**

Competitors

| Name | Where | Launched | Scale | How it differs from this one |
|---|---|---|---|---|
| yihui-dev/awesome-opus5-5-videos | Open source | 2026-09-27 | 13,865★ | Collects the viral videos and prompts made with Opus 5.5 for people to watch; this repo collects tools to install for agents |
| opusvideo/awesome-claude-video and other Opus 5.5 showcase / prompt lists (LeaddeOpenLab, chuspeeism, xianyu110, HiAPIAI) | Open source | 2026-09-25~29 | 53–238★ | All showcases or prompt collections, with no tool directory, grading, or testing |
| ComposioHQ / travisvn awesome-claude-skills | Open source | 2025-10 | 16k–79k★ | Collects every kind of skill, with video only a few lines; this repo covers only video and has collected over four hundred entries |
| heygen-com/hyperframes-community-skills | Open source | 2026-08-25 | 251★ | Official single ecosystem; this repo spans Remotion / HyperFrames / editing / digital humans |
| remotion-dev/skills | Open source | 2026-01-19 | 6,409★ | A single skill, not a directory; included in this repo and recommended in its testing |
| frankxai/awesome-motion-design-agent-skills | Open source | 2026-06-18 | 97★ | Leans toward web motion effects, with no grading, testing, or promotion |
| tavily-api/awesome-hyperframes | Open source | 2026-09-21 | 16★ | Covers only HyperFrames, with little content |
| Skill sites like awesomeclaude.ai / skillsllm.com / skillhey.com | Web | Unknown | Not verified | Web directories or guides that don't accumulate GitHub stars, and rarely grade repos individually |
| Anthropic (official) | Big-company features | — | — | No official video skills directory exists; this repo's testing shows qualified videos can be made without any skill installed |

Where users come from

- **Verified** **Author's X long post** : on 2026-09-27 14:25 UTC @GoSailGlobal published "I open-sourced and categorized all 180 open-source skills that let Claude make videos"; the next day +147

- **Verified** **Chinese AI big-account reposts** : within an hour of the post, @canghe, @lxfater (about 109k followers), and @zstmfhy followed up with recommendations

- **Verified** **The author kept posting** : 9-28 posted "204 video skills packaged" and 10-21 posted the "16 skills tested" tweet, both with links to the site and GitHub

- Inferred **GitHub / Google search** : searching 'claude video skills' on GitHub ranks this first by stars (second place 17★); the companion page 'Claude Video Editing & Remotion Skills: 431 Graded (2026)' appears near the top of search results; after 10-25, daily star growth stayed steady with no spikes

- Inferred **Global 'Claude video' content wave and aggregator sites in late October** : LinkedIn and blogs filled with 'N GitHub repos' posts, getkeyframes cited this repo's testing; skillsllm, mrkeyoor, reveneau, Zhihu, Bilibili, and CSDN have dedicated pages

- **Verified** **A positive loop from skill-author submissions** : 12 issues, mostly skill authors submitting 'Add: my repo'

- **Verified** **Submissions to large awesome lists** : 11-01 submitted #3396 to hesreallyhim/awesome-claude-code, not merged; the author has placed dozens of submissions for other repos

## zhuyansen/awesome-claude-video-skills: why it took off

> Research date 2026-11-06. Data sources: data card, our GitHub repo (repo_profile), GitHub API, Trendshift, X (read via Jina), Exa / Jina search. For the past few days GitHub returns no stargazer details for any repo (neither REST nor GraphQL), so we can't verify who starred it or how legitimate those accounts are.

### One line

This is an awesome list of open-source skills and tools that let coding agents like Claude Code / Codex make videos. It's sorted into categories, has a safety grade for each entry, is bilingual (Chinese and English), and comes with a filterable web page (agentskillshub.top/best/claude-video-skills/). 2026-09-22 Opus 5.5 launched and set off a wave of "making videos with Opus 5.5". On day 5 of the buzz, the author posted a long article on X in the Chinese AI circle, several big accounts reposted it that night, and the repo took off. By late October, "making videos with Claude" was a global content trend, and the repo got a second, steady wave of growth from GitHub and Google search. The author works in SEO / brand promotion (GoSail Lab), and an automated skills-directory site provides the data behind the list.

### 1. Growth data

- Repo created 2026-09-27 04:06 UTC. By 11-06 it had 1,617 stars (data card). The GitHub API showed 1,667 stars and 157 forks that day. Only 2 contributors, and 0 merged PRs (Trendshift).

- Daily star gains (Trendshift [trendshift.io/repositories/265184](https://trendshift.io/repositories/265184)):
    - 9/27 +30, **9/28 +147,9/29 +108** , the launch peak;
    - 9/30–10/24 about +12~35 per day, a plateau;
    - **10/25 +81,10/26 +72** , after that about +35~84 per day, entering a second accelerating phase;
    - Data card: +404 over the last 7 days, +398 over the first 7 days, peak single day 81 (10-27, by our repo's count).

- Our repo first observed it on 10-09, when it had 489 stars. The suspect field is empty and the status is "steady growth".

- The repo was less than 60 days old, so it wasn't on GitHub trending. None of this growth came from trending.

### 2. Where users come from

| Channel | Evidence | Confirmed |
|---|---|---|
| Author's Chinese long-form post on X (X Article) | 2026-09-27 14:25 UTC, published "107 hidden gems with under 50 stars: I categorized and open-sourced all 180 open-source skills that let Claude make videos" (https://x.com/GoSailGlobal/article/2104215886429331580). It mentioned a cost of about 200 USD and reading each README one by one with the Jev model, among other details. The peak of +147 came the day after the post. | Yes |
| Reposts from big Chinese AI accounts | Within half an hour to an hour of the post going up, canghe @canghe reposted it ("This is awesome, bookmarking to study!"). lxfater @lxfater (about 109k followers) wrote a long thread and picked 4 recommendations from the list. @zstmfhy posted "39 video-style skills." All of these went out the same evening (Beijing time) 9-27, which looks like a little circle of people hyping each other up. | Yes |
| Author's own follow-up posts | 9-28 posted "204 video-production skills bundled for use" (status/2105246504034386346). 10-21 posted a hands-on test of "16 skills making the same 15-second promo video" (status/2112801857144029440), with links to both the web page and the GitHub repo. | Yes |
| Search (GitHub search + Google) | Searching "claude video skills" on GitHub, the repo ranks first by stars, and the second place has only 17 stars. When searching "awesome-claude-video-skills" on Jina, the companion page agentskillshub.top/best/claude-video-skills/ ("431 Graded (2026)") and the GitHub repo rank at the top together. The repo's topics are packed with hot terms like claude-code / remotion / hyperframes / opus-5-5. Growth after 10-25 has been roughly even day to day, with no obvious spikes, which fits the pattern of search traffic. | Partly confirmed (search rankings were tested directly; no data on traffic share) |
| Late-October global "making videos with Claude" content wave | From 10-20 to 11-06, a flood of LinkedIn posts and blogs appeared with titles like "N GitHub repos that turn Claude into a video studio" (Rananjay Raj 10-20, Migavel 10-22, Alex Evdokimov 10-31, getkeyframes 11-01, aifire 11-06, and others). Most of them promoted specific skills, but getkeyframes explicitly cited this repo 10-11 as a hands-on reference. Another group of aggregator sites and machine-generated articles built dedicated pages for it: skillsllm, mrkeyoor, reveneau, GitCode/CSDN blogs, Zhihu (JeecgBoot column), Bilibili. | Partly confirmed (shows demand is rising; we don't know how many stars each post brought) |
| Entries sending traffic back | The repo has 12 issues, mostly "Add: my repo" submissions from skill authors (SeeCut, showtime, motionboard, synchronicity-labs, and others). Authors of video skills started treating "getting into this list" as a distribution channel, which can create a positive loop. | Yes (the issues exist; the traffic they sent is unknown) |
| Submitting to large awesome lists | The author 11-01 submitted #3396 "[Resource]: Awesome Claude Video Skills" to hesreallyhim/awesome-claude-code (57k stars). It hasn't been merged yet. For other repos (new-term-radar, backlink-guard) he has run large-scale submission campaigns: from 10-19 to 10-27 he submitted dozens of PRs and issues to awesome-seo, awesome-mcp-servers, ruanyf/weekly, HelloGitHub, mcp.so, raycast and others. The video repo itself got few submissions | Yes (the submissions exist; their effect on the video repo has been small) |
| Timing: Opus 5.5 launch | 9-22 Opus 5.5 launched (@RLanceMartin "Opus 5.5 is out!"), and "Opus 5.5 makes motion-graphics videos" immediately flooded timelines. yihui-dev/awesome-opus5-5-videos was created the same day (9-27) and now has 13.9k stars. | Yes |

### 3. Competitors

| Competitors | Where | Launched | Scale | How it differs |
|---|---|---|---|---|
| yihui-dev/awesome-opus5-5-videos | Open source | 2026-09-27 | 13,865★ | That list collects videos made by Opus 5.5 and their matching prompts, aimed at people who want to watch them. This repo collects tools and skills, aimed at agents that need equipping. Both rode the same wave, but they do different things. |
| opusvideo/awesome-claude-video, LeaddeOpenLab, chuspeeism, xianyu110, HiAPIAI, Li-Evan, and other Opus 5.5 video / prompt lists | Open source | 2026-09-25~29 | 36–238★ | These are also collections of works or prompts. None of them offers a "tool directory + safety grades + hands-on testing". |
| zhuyansen/awesome-opus-5.5-video (the author's own) | Open source | 2026-09-27 | 307★ | A works collection the same author made on the same day. It has one-fifth the stars of this repo, which suggests that for this author, demand for "tools" outweighs demand for "works". |
| ComposioHQ / travisvn / BehiSecc awesome-claude-skills | Open source | 2025-10 | 10k–79k★ | They cover every kind of skill, and video is only a few lines among them. This repo covers only video, and that single category alone has over four hundred entries. |
| heygen-com/hyperframes-community-skills | Open source (official) | 2026-08-25 | 251★ | Covers only the HyperFrames ecosystem. This repo spans Remotion / HyperFrames / editing / digital humans. |
| remotion-dev/skills | Open source (official) | 2026-01-19 | 6,409★ | A single skill, not a directory. This repo includes it and ranks it third in hands-on testing. |
| frankxai/awesome-motion-design-agent-skills | Open source | 2026-06-18 | 97★ | Started earlier, but focuses on web motion (GSAP / Lottie / Rive). No grading or hands-on testing, and no promotion. |
| tavily-api/awesome-hyperframes, Markjinli/awesome-hyperframes-prompt | Open source | 2026-05~09 | 16–20★ | Covers only HyperFrames, and has little content. |
| tuzhechen2005/opus-video-skills, op7418/guizang-product-video-skill, cclank/lanshu-awesome-ai-video-kit | Open source | 2026-05~09 | 445–907★ | Each is "a set of skills" itself, not a third-party directory. |
| awesomeclaude.ai/claude-motion-graphics, skillsllm.com, skillhey.com and other skill sites | Web | Unknown | Not verified | Web directories or guides. They have no GitHub stars or forks, and rarely grade individual repos. |
| Built into Claude / Anthropic | Big-company features | — | — | Anthropic has no official video skills directory. This repo's own hands-on testing even admits that "Claude Code without any skills can still produce a decent video in 4.4 minutes". |

Bottom line: in the niche of "video tool directories for agents", it is the first, and the only one to reach four-digit stars (searching "claude video skills" on GitHub, the second place has only 17★).

### 4. Why it took off

1. **Hit the platform event on day 5** . Opus 5.5 9-22 launched, "Opus makes videos" flooded timelines within two or three days, and people who saw the showcase videos next wanted to know "which tool should I install?" Others collected works and prompts. This one collected tools and caught that demand.

2. **Had a megaphone in the Chinese AI X circle** . The author @GoSailGlobal posted 14.6k tweets and works in brand promotion. Within an hour of the long article, 苍何 and 铁锤人 (109k followers) followed up, and the next day the repo gained +147 / +108.

3. **Deeper content than a typical awesome list** . Each repo's README was read individually, safety-graded (178 SAFE / 2 CAUTION), sorted into 10 categories, and given preview GIFs. 10-11 then added a hands-on test of "16 skills cutting the same film" with recommendations. In tweets and blogs, a "hands-on test" gets shared more readily than a "list".

4. **Automation kept maintenance costs low** . Behind it is the agentskillshub.top data pipeline (claimed to index 260k+ skills and MCP servers, refreshed every 8 hours). The README is rewritten automatically with a "Daily update" every day, and the entry count grew from 168 to 427–431. The repo stays "updated", which helps both search and repeat visits.

5. **The second wave came from search and the global content wave** . Starting around 10-25, English-language content about "making videos with Claude" multiplied. The repo's topics and description cover keywords like "claude video skills / remotion / hyperframes", and the companion web page ranks for search terms like "Claude Video Editing & Remotion Skills". So without any new big-account reposts, it kept gaining +50~80 a day.

6. **The niche was the right pick; the template wasn't the reason** . The author used the same template to batch-create 20 more awesome lists (codex, design, seo, humanizer, obsidian, image, telegram bots, hooks…). ai-reverse-engineering got 60★ and ppt got 41★, and most of the rest landed at 0–11★. Only the video one took off, because it caught the Opus 5.5 video wave.

### 5. Differentiation

- Compared with "works / prompt collections": it collects **tools you can install for agents** .

- Compared with the general awesome-claude-skills: it **covers only video, and covers it deeply** (400+ entries).

- Compared with official ecosystems: it **spans ecosystems** , and also offers **safety grades and same-task hands-on tests** . The conclusions are candid, even stating that "you can pass without a skill", which adds credibility.

- Compared with web skill sites: it **lives on GitHub** , where it can accumulate stars and forks, and it has a web page to capture SEO, so it covers both.

Differentiation is a necessary condition for taking off: if it were just another Opus video collection, it would drown among six or seven similar ones from the same period. But the biggest push was still timing (Opus 5.5) plus the author's X channel. Differentiation determines that it "sticks around and keeps getting cited," not why it "exploded on day one."

### 6. How visible it is

It spread in small circles. Chinese AI X circles (Cang He, Tie Chui Ren, etc.) and reposts and machine-generated articles on Zhihu, Bilibili, CSDN and GitCode wrote about it. On the English side, only aggregator sites (skillsllm, mrkeyoor, reveneau) and a few blogs (getkeyframes) cited it; I found no English big account or media outlet writing about it specifically. The author, who makes open-source maps to drive traffic to agentskillshub.top, has no revenue problem and has never disclosed revenue.

### 7. Copycats

- UnderleveledBuilder/awesome-short-drama-skills (2026-10-13 built, 1,901★): same approach, "curated sub-category skills + Chinese + GitHub Actions auto-generating candidate projects," transplanted to AI short drama / comic dramas. It also submitted to hesreallyhim. Whether it is copying this one is unverified.

- 1368546410qq/dsh-video-skills (10-29,2★): a DSH plugin for "browse 373 video skills and install," close in scale to this repo, apparently using its data.

- frankxai/awesome-ai-video-routes (10-20,0★): "every path to make video with Claude."

- A large batch of same-named xxx/claude-video-skills repos (mostly 0–17★); most are skill collections, not directories.

- The author is also copying themselves: awesome-codex-ppt-skills, awesome-claude-seo-skills and 20 more lists from the same template.

### 8. What others can learn

- **After a big event, build the directory for the "next question"** . In a frenzy, everyone collects works first; next someone will ask "what tools do I use to make it?" Building a directory at this layer faces far less competition.

- **A list can't just be a list** : safety grades, categories, preview GIFs, plus one "same-topic live test" give people a story to pass along when they share it.

- **Run a GitHub repo and a own website in parallel** : the repo gets stars and GitHub in-site search; the site gets Google keywords (titles like "431 Graded (2026)"), and the two link to each other.

- **Auto-update the README daily** to keep a "still alive" signal; turn "Add: xxx" issues into a distribution entry point for skill authors, forming a positive loop.

- **The launch relied on a big account in the familiar circle following up within an hour** ; the first-day peak set the starting point for later search rankings.

### 9. Uncertain

- Stargazer-level detail isn't available, so star farming or mutual liking can't be ruled out; still, the curve's shape (event peak → plateau → steady climb) looks like organic growth.

- I didn't find which piece of content triggered 10-25's second acceleration; it may be several English posts stacking, plus rising search rankings.

- Follower counts and repost data for accounts like Cang He are incomplete (X pages require login).

- There is no data on agentskillshub.top's actual traffic or how much the repo drives traffic to the site.

- Once hesreallyhim/awesome-claude-code's submission (#3396) is merged, it could bring another wave.

### Sources

- Data card /home/ljr/fanout/why/cards/gh_zhuyansen__awesome-claude-video-skills.json

- GitHub: [github.com/zhuyansen/awesome-claude-video-skills](https://github.com/zhuyansen/awesome-claude-video-skills) (README, commits, issues); all of the author's repos [github.com/zhuyansen](https://github.com/zhuyansen)

- Trendshift:[trendshift.io/repositories/265184](https://trendshift.io/repositories/265184)

- X:[x.com/GoSailGlobal/article/2104215886429331580](https://x.com/GoSailGlobal/article/2104215886429331580) ;[x.com/canghe/status/2104223530527412418](https://x.com/canghe/status/2104223530527412418) ;[x.com/lxfater/status/2104232636738588797](https://x.com/lxfater/status/2104232636738588797) ;[x.com/GoSailGlobal/status/2105246504034386346](https://x.com/GoSailGlobal/status/2105246504034386346) ;[x.com/GoSailGlobal/status/2112801857144029440](https://x.com/GoSailGlobal/status/2112801857144029440)

- Companion site: [agentskillshub.top/best/claude-video-skills/](https://agentskillshub.top/best/claude-video-skills/)

- Submission: [github.com/hesreallyhim/awesome-claude-code/issues/3396](https://github.com/hesreallyhim/awesome-claude-code/issues/3396)

- Similar: [github.com/yihui-dev/awesome-opus5-5-videos](https://github.com/yihui-dev/awesome-opus5-5-videos), [github.com/UnderleveledBuilder/awesome-short-drama-skills](https://github.com/UnderleveledBuilder/awesome-short-drama-skills), [github.com/frankxai/awesome-motion-design-agent-skills](https://github.com/frankxai/awesome-motion-design-agent-skills)

- Secondary spread: [zhuanlan.zhihu.com/p/2090945134301024754](https://zhuanlan.zhihu.com/p/2090945134301024754), [www.bilibili.com/video/BV17gpP6aEpy/](https://www.bilibili.com/video/BV17gpP6aEpy/), [skillsllm.com/skill/awesome-claude-video-skills](https://skillsllm.com/skill/awesome-claude-video-skills), [getkeyframes.com/blogs/guides/claude-video-skills](https://getkeyframes.com/blogs/guides/claude-video-skills)

- Our GitHub database repo_profile (2026-11-06)

This article was written by AI from public data and the data we collect daily. Numbers are sourced, and inferences are labeled.